Bi-Level Optimization Augmented with Conditional Variational Autoencoder for Autonomous Driving in Dense Traffic
On the map: End-to-end autonomous driving → Reinforcement learning for driving
Abstract (arXiv)
Autonomous driving has a natural bi-level structure. The goal of the upper behavioural layer is to provide appropriate lane change, speeding up, and braking decisions to optimize a given driving task. However, this layer can only indirectly influence the driving efficiency through the lower-level trajectory planner, which takes in the behavioural inputs to produce motion commands. Existing sampling-based approaches do not fully exploit the strong coupling between the behavioural and planning layer. On the other hand, end-to-end Reinforcement Learning (RL) can learn a behavioural layer while incorporating feedback from the lower-level planner. However, purely data-driven approaches often fail in safety metrics in unseen environments. This paper presents a novel alternative; a parameterized bi-level optimization that jointly computes the optimal behavioural decisions and the resulting downstream trajectory. Our approach runs in real-time using a custom GPU-accelerated batch optimizer, and a Conditional Variational Autoencoder learnt warm-start strategy. Extensive simulations show that our approach outperforms state-of-the-art model predictive control and RL approaches in terms of collision rate while being competitive in driving efficiency.
Key facts. Cited by 2 papers in the map (data through 14 Oct 2026). Placed in Reinforcement learning for driving, one of 9 directions in End-to-end autonomous driving. Builds on 13 of its 31 references that are in the map. First posted on arXiv 5 Dec 2022.
Builds on
13 of 31 references.
Closest ancestors · 1
Same direction, newest first
- 2021A Safe Hierarchical Planning Framework for Complex Driving Scenarios based on Reinforcement Learning17 Jan
Other references · 12
- 2015Generating Sentences from a Continuous Spacecited by 612
- 2017OptNet: Differentiable Optimization as a Layer in Neural Networkscited by 599
- 2016Training Neural Networks Without Gradients: A Scalable ADMM Approachcited by 61
- 2019Combining Planning and Deep Reinforcement Learning in Tactical Decision Making for Autonomous Drivingcited by 52
- 2021STORM: An Integrated Framework for Fast Joint-Space Model-Predictive Control for Reactive Manipulationcited by 47
- 2018Automated Speed and Lane Change Decision Making using Deep Reinforcement Learningcited by 36
- 2021Policy Search for Model Predictive Control with Application to Agile Drone Flightcited by 20
- 2013Optimal parameter selection for the alternating direction method of multipliers (ADMM): quadratic problemscited by 19
- 2019Dynamic Input for Deep Reinforcement Learning in Autonomous Drivingcited by 11
- 2021Multi-Modal Model Predictive Control through Batch Non-Holonomic Trajectory Optimization: Application to Highway Drivingcited by 9
- 2018Fast UAV Trajectory Optimization using Bilevel Optimization with Analytical Gradientscited by 5
- 2020Decision-making for Autonomous Vehicles on Highway: Deep Reinforcement Learning with Continuous Action Horizoncited by 2
Built on by
Citing papers per year: 2023: 1; 2024: 1.
2 papers.
Similar, not cited
Close in wording, no citation either way
- 2023End-to-End Learning of Behavioural Inputs for Autonomous Driving in Dense Traffic
- 2023Parameterized Decision-making with Multi-modal Perception for Autonomous Driving
- 2023Efficient Reinforcement Learning for Autonomous Driving with Parameterized Skills and Priors
- 2026DriveAnchor: Progressive Anchor-based Flow Learning for Autonomous Driving Planning
- 2025RoaD: Rollouts as Demonstrations for Closed-Loop Supervised Fine-Tuning of Autonomous Driving Policies
- 2025CuRLA: Curriculum Learning Based Deep Reinforcement Learning for Autonomous Driving
- 2023Boosting Offline Reinforcement Learning for Autonomous Driving with Hierarchical Latent Skills
- 2025A Comprehensive Review of Reinforcement Learning for Autonomous Driving in the CARLA Simulator